Uncertainty Estimation based on Geometric Separation
This work addresses uncertainty estimation for mission-critical applications like autonomous driving, though it appears incremental as it builds on existing post-hoc calibration techniques.
The authors tackled the problem of uncertainty estimation in machine learning by proposing a geometric-based approach that uses the distance of inputs from training data to estimate uncertainty, then calibrates it with post-hoc techniques. They demonstrated that their method achieves more accurate uncertainty estimations than recent approaches across various datasets and models, with optimizations enabling near real-time implementation for large datasets.
In machine learning, accurately predicting the probability that a specific input is correct is crucial for risk management. This process, known as uncertainty (or confidence) estimation, is particularly important in mission-critical applications such as autonomous driving. In this work, we put forward a novel geometric-based approach for improving uncertainty estimations in machine learning models. Our approach involves using the geometric distance of the current input from existing training inputs as a signal for estimating uncertainty, and then calibrating this signal using standard post-hoc techniques. We demonstrate that our method leads to more accurate uncertainty estimations than recently proposed approaches through extensive evaluation on a variety of datasets and models. Additionally, we optimize our approach so that it can be implemented on large datasets in near real-time applications, making it suitable for time-sensitive scenarios.